Genome size is underestimated based on C-value

[joeycurti3]

Hello,

We are generating a genome for the California red-legged frog (Rana draytonii) which has an estimated genome size of around 10GB based on flow cytometry (https://www.genomesize.com/result_species.php?id=342).

We recently ran genomescope on our Hifi data to estimate genome size using k=31 which is recommended in the package release for large genomes >10GB (Vurture et al. 2017). Overall, I think the model fit looks good. However, the genome size estimate is substantially lower than expected (6.3GB; see attached). I'm trying to understand if this is an error in the estimation caused by faulty parameter defaults or if this is a good estimate. Can you advise if there is anything that we can do to evaluate this? Model fit information is also below. Finally, here are the parameters that we generated the histogram file with that was uploaded into Genomescope2:

kmc -k31 -t16 -m64 -ci1 -cs100000 "@${SCRATCH_DIR}/input_files.txt" "${DB_NAME}" "${SCRATCH_DIR}"

Model fit results:

GenomeScope version 2.0
input file = user_uploads/hDPQUFDopJbaCAxCEm6m
output directory = user_data/hDPQUFDopJbaCAxCEm6m
p = 2
k = 31

property min max
Homozygous (aa) 99.4622% 99.5185%
Heterozygous (ab) 0.481482% 0.537793%
Genome Haploid Length 6,282,716,562 bp 6,307,281,925 bp
Genome Repeat Length 3,380,924,983 bp 3,394,144,368 bp
Genome Unique Length 2,901,791,579 bp 2,913,137,557 bp
Model Fit 48.7941% 96.3739%
Read Error Rate 0.115072% 0.115072%

Formula: y_transform ~ x^transform_exp * length * predict2_0(r1, k, d,
kmercov, bias, x)

Parameters:
Estimate Std. Error t value Pr(>|t|)
d 5.674e-02 1.342e-03 42.29 <2e-16
r1 5.096e-03 1.408e-04 36.20 <2e-16
kmercov 1.854e+01 1.809e-02 1025.02 <2e-16
bias 8.633e-01 1.630e-02 52.98 <2e-16
length 3.082e+09 9.104e+06 338.55 <2e-16

Signif. codes: 0 '' 0.001 '' 0.01 '' 0.05 '.' 0.1 ' ' 1

Residual standard error: 38090000 on 1995 degrees of freedom

Number of iterations to convergence: 6
Achieved convergence tolerance: 1.49e-08

[mschatz]

Thanks for your interest. The fit is not too bad, but there is an obvious gap in the model vs observed in the heterozygosity peak so it is probably overestimating the level of heterozygosity. For the genome size, I noticed your kmer coverages drop off at 1e5 (e.g. 100,000) but for a genome of this size you should probably go even larger (1e6 or 1e7) or you will undercount how often the most common repeats occur. This is important for these very large genomes because they are often filled with TEs and/or satellite sequences that are nearly identical so can get kmers that genuinely occur hundreds of thousands or millions of times throughout the genome. Unfortunately you will need to rerun your kmer counter to get the counts this high, but Ive seen this increase the genome size by 10% to 20% so this would help to narrow the gap with your flow data. Also are you using illumina or hifi reads here? For illumina, I would recommend sticking to k=21 just so that you can avoid sequencing errors on the short reads (but for hifi, k=31 is probably better so that more kmers are unique). But for the assembly, you will definitely want to use HiFi or ONT to have any hope to get through the repeats Good luck! Mike

…

On Wed, Jan 14, 2026 at 5:55 PM joeycurti3 ***@***.***> wrote: *joeycurti3* created an issue (schatzlab/genomescope#166) <#166> Hello, We are generating a genome for the California red-legged frog (Rana draytonii) which has an estimated genome size of around 10GB based on flow cytometry (https://www.genomesize.com/result_species.php?id=342). We recently ran genomescope on our Hifi data to estimate genome size using k=31 which is recommended in the package release for large genomes >10GB (Vurture et al. 2017). Overall, I think the model fit looks good. However, the genome size estimate is substantially lower than expected (6.3GB; see attached). I'm trying to understand if this is an error in the estimation caused by faulty parameter defaults or if this is a good estimate. Can you advise if there is anything that we can do to evaluate this? Model fit information is also below. Finally, here are the parameters that we generated the histogram file with that was uploaded into Genomescope2: kmc -k31 -t16 -m64 -ci1 -cs100000 "@${SCRATCH_DIR}/input_files.txt" "${DB_NAME}" "${SCRATCH_DIR}" Model fit results: GenomeScope version 2.0 input file = user_uploads/hDPQUFDopJbaCAxCEm6m output directory = user_data/hDPQUFDopJbaCAxCEm6m p = 2 k = 31 property min max Homozygous (aa) 99.4622% 99.5185% Heterozygous (ab) 0.481482% 0.537793% Genome Haploid Length 6,282,716,562 bp 6,307,281,925 bp Genome Repeat Length 3,380,924,983 bp 3,394,144,368 bp Genome Unique Length 2,901,791,579 bp 2,913,137,557 bp Model Fit 48.7941% 96.3739% Read Error Rate 0.115072% 0.115072% Formula: y_transform ~ x^transform_exp * length * predict2_0(r1, k, d, kmercov, bias, x) Parameters: Estimate Std. Error t value Pr(>|t|) d 5.674e-02 1.342e-03 42.29 <2e-16 r1 5.096e-03 1.408e-04 36.20 <2e-16 kmercov 1.854e+01 1.809e-02 1025.02 <2e-16 bias 8.633e-01 1.630e-02 52.98 <2e-16 length 3.082e+09 9.104e+06 338.55 <2e-16 Signif. codes: 0 '*' 0.001 '' 0.01 '*' 0.05 '.' 0.1 ' ' 1 Residual standard error: 38090000 on 1995 degrees of freedom Number of iterations to convergence: 6 Achieved convergence tolerance: 1.49e-08 image.19.png (view on web) <https://github.com/user-attachments/assets/2f454037-da12-4621-af46-d43fb4cab6f1> — Reply to this email directly, view it on GitHub <#166>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AABP346X5TQ7BOQO77IORZL4G3CNNAVCNFSM6AAAAACRXITU42VHI2DSMVQWIX3LMV43ASLTON2WKOZTHAYTKMRSGY3DGNQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>